Multi Colony Collaborative Algorithm And Its Application Evolution

Artificial Bee Colony (ABC) algorithm is an emerging intelligent algorithm in SwarmIntelligence. The inspiration of the ABC algorithm is derived from group behaviors of bees innature. This algorithm adopts the mechanisms of changing roles and selecting multi-leaders tosolve optimization problems, which has simple parameters and better performance onconvergence compared with the other algorithms in SI. Therefore, it has been focused greatlyrecently and its application scope has been extended from the numerical optimization to manyother areas such as engineering optimization, machine learning, computer graphics and datamining.However, the ABC algorithm is insufficient to extend the diversity of solutions and may betrapped into local extreme points at the end of the evolutionary process when solving complexoptimization problems, especially multimodal optimization problems. It will decrease theperformance of searching the optimum solutions largely. Therefore, this paper improves theABC algorithm, changing the evolutionary strategy based on single-species and adopting themechanism of multi-species co-evolution to extend the diversity of solutions in order toguarantee avoidance of local extreme points. Moreover, several groups of experiments are madeon the ABC algorithm based on multi-species co-evolution (Co-ABC) to test its performance.The problem of crowd evacuation simulation is also taken as an application example todemonstrate the Co-ABC algorithm’s practical value and effectiveness.This paper’s main content and innovations are listed below.(1) In this paper, we combine the ABC algorithm with multi-species co-evolutionmechanism, proposing three versions of the communication model, including the communicationmodel based on central control (CCM), the communication model based on annular transmission(TCM) and the integrated model (ICM). The CCM uses a global communication method toaccelerate convergence, while it is sensitive to the local optimum. The TCM uses a localcommunication method, which has slower convergence but higher diversity of the solutions. TheICM combines the convergence superiority in the CCM with the accuracy superiority in theTCM in order to obtain complementary advantages, improving performances of accuracy,convergence and robustness at the same time.(2) Propose a method to shrink the solution space and species based on the maximumlikelihood estimation and employed norm. In the process of solving complex optimizationproblems by using the Co-ABC algorithm, the candidate solution space is shrinked and refinedgradually to the global optimization area by evaluating the employed norm. The scale andconstitution of each species are also adjusted in the meanwhile. Each species decreases the scaleproportionately to the shrinkage of the solution space, reserving the elites and eliminating the solutions with low fitness. The computational complexity can be reduced and the efficiency ofsolving optimization problems can be improved by taking advantage of shrinking solution spaceand decreasing the scale of species.(3) This paper takes the problem of crowd evacuation simulation as an application exampleof the Co-ABC algorithm, proposing a crowd evacuation simulation method based on theCo-ABC algorithm. In this method, the Co-ABC algorithm based on the ICM is used to planpath in order to generate the optimal path without collision from the initial position to the targetsafety area. Compared with the traditional path-planning algorithms, this method has simplermodel and parameters. It is also insensitive to the size and position of obstacle, which guaranteesthat it can be transformed flexibly from the simple environment to the complex environment.This application example not only demonstrates the practical value of the Co-ABC algorithm,but also provides a fresh idea for the research on crowd evacuation, which can improve theefficiency of simulation’s arrangement and implementation.(4) According to the theory above, relating scientific projects under research, a crowdevacuation simulation system based on the Co-ABC algorithm is developed and implemented inorder to simulate the evacuation process of crowd. The sysytem provides six main functions,including environment modeling, parameter setting, hazard and safe area setting, destinationsetting, path planning and data output. Users can set the simulation environment and group sizeflexibly in order to simulate the evacuation process in many different environments. Thesimulation results demonstrate that the system can generate the evacuation solution with rapid,orderly features and uniform distribution, exhibiting good verisimilitude and visual effect.